Objective:

To evaluate a smartphone-based AI platform for the early detection of ocular surface malignancies.

Approach:

• Development of CaptureTumor (CaT): An AI-powered screening system combining smartphone photography, automated image analysis, public health outreach, and specialist referral pathways.
• Training and Deployment: The deep-learning system was trained on over a decade of slit-lamp photographs and adapted for smartphone use, deployed as a WeChat Mini Program.
• Public Engagement: The screening campaign reached over 256,000 people, with 13,243 accessing the application and 614 completing self-screening.
• Performance Metrics: The system achieved an AUC of 0.977 in real-world screening, with sensitivity of 89.3% and specificity of 95.9%.

Key Findings:

• The smartphone-based system identified 20 histopathologically confirmed malignancies, including 14 basal cell carcinomas and six malignant melanomas.
• Nineteen cases were previously undiagnosed, indicating earlier detection.
• Patients identified through the app required fewer referrals to access specialist services compared to historical averages.

Interpretation:

The study highlights the role of consumer technologies in enhancing ocular malignancy screening and improving access to care.

Limitations:

• The study's findings are based on a specific population in China and may not be generalizable.
• The self-screening process may have selection bias, as not all users may represent the broader population.

Conclusion:

The integration of AI-enabled self-screening with public health outreach may improve early detection of ocular malignancies.

Sources:

• JAMA Ophthalmology